System and method for controlling a pipeline polishing robot

Abstract

The present invention discloses a system and method for controlling a pipeline polishing robot, relating to the field of robot control technology. The system is composed of several functional modules, including: a variable diameter adaptive preprocessing module, which obtains a pipe diameter curve through continuous laser scanning, and splices the continuously scanned point cloud data in real time through a SLAM algorithm to construct a dynamic three-dimensional topological model of the pipeline; obtains the robot's posture data and calculates the offset of the robot's central axis in the pipeline; obtains the monitored ambient temperature in the pipeline, constructs a dynamic thermal expansion model of the pipeline through heat conduction finite element analysis, and corrects the pipe diameter value according to the thermal expansion model; calculates the expansion and contraction of the umbrella structure through a fuzzy PID controller based on the offset and the corrected pipe diameter value, and generates an umbrella structure expansion and contraction instruction; a dust collection head is synchronously unfolded to the pipe wall; and a multi-process collaborative control module.

Description

Technical Field

[0001] The present invention relates to the technical field of robot control, and in particular to a system and method for controlling a pipeline polishing robot. Background Art

[0002] As the water transfer tunnel is a key water transfer structure in a water conservancy project, its structural system consists of three major parts: the water inlet gate, the main tunnel section and the terminal control chamber, forming a complete water flow transmission channel; the water inlet gate is responsible for water level regulation and flow control, and adopts a reinforced concrete frame structure and is equipped with a two-way opening and closing gate; the main tunnel section is the core water transfer carrier of the tunnel, and adopts a Q345B low-alloy steel plate lining structure with a wall thickness of 12mm and a pipe diameter range of DN1200-DN1300; the terminal chamber section is equipped with an arc-shaped steel working gate and a hydraulic opening and closing system to achieve precise flow control; during continuous operation, this structural system has been affected by multiple factors such as hydraulic loads and environmental erosion, and has shown serious material performance degradation.

[0003] Traditional anti-corrosion repair mostly relies on manual entry into the pipeline for operation. The tunnel entrance is located at the bottom of the reservoir, and construction workers can only enter and exit through a single-side exit. The ventilation conditions are extremely poor, and working in a confined space can easily cause accidents such as suffocation and poisoning. The manual rust removal speed is only 0.5-1㎡ / h, and is limited by the pipe diameter. It is usually only suitable for manhole entry and exit above DN800, and a 281m pipeline requires more than 30 days. The thickness of the manually sprayed coating fluctuates by >15%, and it is easy to miss the coating and accumulate in the variable diameter section, resulting in local anti-corrosion failure. Most commercially available robots are designed for small diameter pipes below DN600, and can only perform local corrosion protection of welds. They cannot meet the continuous operation requirements of large diameters (>DN1000). In addition, traditional equipment lacks the ability to adapt to pipe diameters. Manual intervention is required to adjust the mechanical structure in the variable diameter area, interrupting construction continuity. Summary of the Invention

[0004] To achieve the above objectives, the present invention is implemented through the following technical solutions:

[0005] A system for controlling a pipe polishing robot, comprising:

[0006] The variable diameter adaptive preprocessing module acquires the pipe diameter curve through continuous laser scanning, and uses the SLAM algorithm to stitch the continuously scanned point cloud data in real time to construct a dynamic three-dimensional topological model of the pipeline. It also acquires the robot's posture data and calculates the offset of the robot's central axis within the pipeline. It obtains the monitored ambient temperature within the pipeline and constructs a dynamic thermal expansion model of the pipeline through heat conduction finite element analysis. The pipe diameter value is corrected according to the thermal expansion model. Based on the offset and the corrected pipe diameter value, the fuzzy PID controller calculates the expansion and contraction of the umbrella structure and generates the expansion and contraction instructions for the umbrella structure.

[0007] The multi-process collaborative control module dynamically sets the spraying baseline speed and relationship based on the pipe diameter curve and coating thickness requirements. It uses the YOLOv5 model to detect foreign objects on the pipe wall in real time. When the obstacle height is greater than the height threshold c, it pauses the process, triggers removal, and continues the process. If removal is still not possible, the location is marked and the control console is notified for manual intervention.

[0008] When the obstacle height is identified to be less than or equal to the height threshold c, the vehicle is stopped and local grinding is triggered;

[0009] The coating thickness closed-loop control module obtains thickness feedback in real time, compares it with the target value, adjusts the spray motor speed, and performs floating compensation on the nozzle.

[0010] Furthermore, the process of obtaining the pipe diameter curve is as follows:

[0011] Perform arc fitting on the spliced ​​point cloud data to calculate the inner diameter value of each cross section; identify the diameter-changing area through curvature calculation and generate a segmented continuous pipe diameter curve.

[0012] Furthermore, the process of real-time splicing by SLAM algorithm is as follows:

[0013] Using the LIO-SLAM algorithm, dynamic stitching is achieved through the following steps:

[0014] S101: Use IMU data, accelerometer and gyroscope to calculate the robot's posture changes in real time;

[0015] S102: Aligning point cloud data of adjacent frames using the ICP algorithm;

[0016] S103: Based on the SLAM framework, the spliced ​​point cloud data is constructed into a dynamic 3D topological model of the pipeline;

[0017] Voxel filtering is used to reduce the point cloud density while preserving geometric features; statistical outlier detection is used to remove noise points.

[0018] Furthermore, the process of constructing the dynamic three-dimensional topological model of the pipeline is:

[0019] Based on dense point cloud data, the entire pipe cross section is covered to generate the inner wall of the complex shape of the pipe. Each frame of point cloud data contains spatial point coordinate information, and the point cloud is filtered to remove isolated points.

[0020] Using the ICP algorithm, a newly acquired point cloud frame is matched with the previously constructed map to find the rigid body transformation between the two frames of point clouds. By integrating the results of multiple local registrations, a global map is formed. Based on the registered point cloud data, a computer vision algorithm is used to extract the key geometric features inside the pipeline and dynamically update them.

[0021] Furthermore, the process of calculating the offset of the central axis of the robot in the pipeline is:

[0022] The robot's angular velocity and acceleration are acquired through sensors, and the central axis of the pipeline is extracted as the ideal trajectory based on the three-dimensional topological model of the pipeline constructed by laser scanning. The actual central axis position of the robot is calculated by combining the attitude angle data of the IMU and the robot's kinematic model. The attitude angle measured by the IMU is converted into the position in the robot's coordinate system, and the robot's translational displacement is acquired through the odometer. The actual central axis position is compared with the ideal trajectory to calculate the deviation.

[0023] Furthermore, the process of correcting the pipe diameter value is as follows:

[0024] Based on the material properties of the pipeline, the model is meshed, discretizing the continuous pipeline structure into many small cells. Appropriate boundary conditions are set according to the actual situation, and the finite element method is used to solve the heat conduction equation to predict the temperature field distribution of each part of the pipeline over time under given conditions.

[0025] According to the temperature field distribution, the thermal expansion of each position of the pipeline is calculated section by section; the average value of the inner and outer wall temperatures is used as the correction basis.

[0026] Furthermore, the specific process of generating the umbrella structure expansion instruction is as follows:

[0027] The three parameters of traditional PID, namely proportional gain Kp, integral gain Ki, and differential gain Kd, are dynamically adjusted through fuzzy logic. The input is the error and the error change rate, and the output is the PID parameter correction value. Mamdani fuzzy reasoning is used to calculate the weight of each rule, and the center of gravity method is used to convert the fuzzy output into accurate Kp, Ki, and Kd values. The expansion and contraction amount is adjusted in real time based on the corrected Kp, Ki, and Kd values.

[0028] Furthermore, the process of dynamically setting the spraying reference speed and relationship is as follows:

[0029] By determining the spraying baseline speed and based on the principle of paint flow conservation, precise control is achieved for the dynamic balance between pipe diameter, coating thickness, and paint flow. Laser scanning is used to obtain the real-time pipe diameter, which, combined with temperature sensor data, calculates the thermal expansion of the pipeline to obtain the corrected actual pipe diameter. The effective flow rate is adjusted according to the ambient temperature to offset changes in paint fluidity caused by high temperatures.

[0030] Set the spraying speed according to the output efficiency of the spraying equipment under different conditions.

[0031] Furthermore, the process of floating compensation of the nozzle is as follows:

[0032] Electromagnetic eddy currents are used to automatically detect coating thickness, and the operator sets the target coating thickness through the HMI. The motor speed is dynamically adjusted based on the deviation to eliminate long-term deviations. Integrated force sensors and displacement sensors detect the contact force and relative position between the printhead and the workpiece in real time. The floating resistance of the printhead is automatically adjusted based on the contact force, and the printhead angle is adjusted through floating compensation.

[0033] A method for controlling a pipe polishing robot comprises the following steps:

[0034] Step 1: Obtain the pipe diameter curve through continuous laser scanning, and use the SLAM algorithm to stitch the continuously scanned point cloud data in real time to construct a dynamic three-dimensional topological model of the pipeline; obtain the robot's posture data and calculate the offset of the robot's central axis in the pipeline; obtain the monitored ambient temperature in the pipeline, and use heat conduction finite element analysis to construct a dynamic thermal expansion model of the pipeline, and correct the pipe diameter value according to the thermal expansion model; based on the offset and the corrected pipe diameter value, calculate the expansion and contraction of the umbrella structure through a fuzzy PID controller, and generate the expansion and contraction instructions of the umbrella structure;

[0035] Step 2: Dynamically set the spraying baseline speed and relationship based on the pipe diameter curve and coating thickness requirements; use the YOLOv5 model to detect foreign objects on the pipe wall in real time. When the obstacle height is greater than the height threshold c, the process is paused, the removal is triggered, and the process continues; if the obstacle still cannot be removed, the location is marked and the control console is notified for manual intervention;

[0036] When the obstacle height is identified to be less than or equal to the height threshold c, the vehicle is stopped and local grinding is triggered;

[0037] Step 3: Get thickness feedback in real time, compare it with the target value, adjust the spray motor speed, and perform floating compensation on the nozzle.

[0038] The system and method for controlling a pipe polishing robot provided by the present invention have the following beneficial effects:

[0039] (1) The present invention dynamically adjusts the expansion and contraction of the umbrella structure through a fuzzy PID controller, and combines the real-time offset and the corrected tube diameter value to achieve dynamic compensation with millimeter-level accuracy. The fuzzy PID controller dynamically adjusts the Kp / Ki / Kd parameters through fuzzy logic to suppress sensor noise and external interference. In harsh environments such as dust and vibration, the system can still operate stably.

[0040] (2) The present invention uses a fuzzy rule base to automatically optimize PID parameters, eliminating the need for manual repeated tuning and shortening debugging time. The telescopic range of the umbrella structure is dynamically adjusted by correcting the pipe diameter value, supporting a variety of pipe specifications. The same device can be adapted to pipes of different diameters without the need to replace hardware. The closed-loop feedback mechanism is combined with the fuzzy PID controller to ensure that the nozzle / sensor is always in the optimal working position. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] Figure 1 Schematic diagram of the system flow of the present invention;

[0042] Figure 2 Schematic diagram of the overall method of the present invention. DETAILED DESCRIPTION

[0043] The following will provide a clear and complete description of the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0044] Example 1

[0045] See also Figure 1 , Embodiment 1 of the present application provides a system for controlling a pipe polishing robot, the system comprising:

[0046] The variable diameter adaptive preprocessing module acquires the pipe diameter curve through continuous laser scanning, and stitches the continuously scanned point cloud data in real time using the SLAM algorithm to construct a dynamic three-dimensional topological model of the pipeline. It also acquires the robot's posture data and calculates the offset of the robot's central axis within the pipeline. It obtains the monitored ambient temperature within the pipeline and constructs a dynamic thermal expansion model of the pipeline through heat conduction finite element analysis. The pipe diameter value is corrected according to the thermal expansion model. Based on the offset and the corrected pipe diameter value, the fuzzy PID controller calculates the multi-degree-of-freedom expansion and contraction of the umbrella structure and generates an expansion and contraction instruction for the umbrella structure. The vacuum head is simultaneously deployed to the pipe wall.

[0047] The robot is 6.8 meters long and consists of a power tractor, a grinding and rust removal vehicle, a vacuum cleaning vehicle, a spray anti-corrosion vehicle, a wireless remote control module, and a control box.

[0048] Power tractor: Built-in traction motor, working lithium battery pack, wireless bridge module, LED lights, system switches, cameras, laser scanners and other sensors are arranged on the front of the vehicle;

[0049] Grinding and rust removal vehicle: This is the key improved part of the robot. It innovatively uses an umbrella-shaped radial grinding structure that can operate continuously. The central spindle supports the entire grinding component, enabling it to rotate 360°.

[0050] Vacuum cleaning vehicle: built-in sealed dust collection bin, telescopic and rotating vacuum head;

[0051] Spraying vehicle: It is equipped with two sets of quantitative paint pumps and two silos. A high-speed spraying motor is installed at the front end of the quantitative pump. The sprayer is fixed at the end of the motor shaft. The paint and curing agent enter the sprayer at the same time through the feeding copper tube, are fully mixed on the cone surface of the sprayer, and then sprayed out at high speed by the sprayer.

[0052] Wireless module and control box: The wireless module uses a pair of high-power network bridges, one installed on the battery pack tractor and the other connected to the control box; the control box cover has a display and touch screen and a built-in control system.

[0053] Get the pipe diameter curve:

[0054] A multi-line laser radar is used with a scanning resolution of 0.1mm, acquiring 100 frames of point cloud data per second and a ranging range of 30-500mm. It is installed on the central axis of the robot body and driven by a servo motor to rotate and scan 360°. If the pipe diameter is large, multiple sets of laser sensors can be deployed to scan synchronously from different angles to avoid occlusion problems. Arc fitting is performed on the spliced ​​point cloud data to calculate the inner diameter value of each cross section. The spliced ​​point cloud data is denoised and plane segmented to extract the contour points of the pipe cross section. The least squares method or RANSAC algorithm is used to fit the contour points to the arc and solve the optimal center and radius. The inner diameter value of each cross section is obtained based on the fitting results. Multiple cross-section point sets are cut out along the pipe axis from the spliced ​​and corrected three-dimensional point cloud data. Two-dimensional arc fitting is performed on the contour points of each cross section to obtain the optimal center and radius of the cross section. Double the fitting radius is used as the inner diameter value of the cross section to form pipe diameter distribution data for each cross section, and pipe diameter change trend analysis and diameter change area identification are performed.

[0055] Real-time stitching through SLAM algorithm:

[0056] Using the LIO-SLAM algorithm, dynamic stitching is achieved through the following steps:

[0057] S101: Use IMU data, accelerometer and gyroscope to calculate the robot's posture changes in real time;

[0058] S102: Align the point cloud data of adjacent frames using the ICP algorithm to eliminate motion distortion;

[0059] S103: Based on the SLAM framework, the spliced ​​point cloud data is constructed into a dynamic 3D topological model of the pipeline;

[0060] Voxel filtering is used to reduce the point cloud density while preserving geometric features; statistical outlier detection is used to remove noise points;

[0061] Construct a dynamic 3D topology model of the pipeline:

[0062] Based on dense point cloud data, the entire pipeline cross-section is covered, and the complex inner wall of the pipeline is generated in detail. Each frame of point cloud data contains spatial point coordinate information relative to the LiDAR coordinate system, reflecting the geometric shape of the pipeline inner wall at a specific moment. Due to environmental factors such as dust and water vapor, the raw point cloud data may contain noise. The point cloud is filtered to remove isolated points or outliers, improving the accuracy of subsequent processing.

[0063] Using the ICP algorithm, a newly acquired point cloud frame is matched with the previously constructed map to find the optimal rigid body transformation between the two point clouds, making them as close as possible to each other. By integrating the results of multiple local registrations, a consistent global map is formed, solving the cumulative error and improving the quality of the 3D model.

[0064] Based on the registered point cloud data, computer vision algorithms are used to extract key geometric features within the pipeline, such as diameter changes, curvature, and branching, to build a detailed topological structure. As the robot advances, new point cloud data is continuously added to the existing map, enabling dynamic updates to the pipeline 3D model and ensuring high mapping accuracy even during long-distance exploration.

[0065] Calculate the offset of the robot from the center axis in the pipe:

[0066] By installing a three-axis gyroscope and a three-axis accelerometer, the angular velocity and acceleration of the robot body are measured;

[0067] Gyroscope: provides angular velocity data, which can be used to calculate attitude angle through integration;

[0068] Accelerometer: Calculates static attitude angles, such as pitch and roll angles, based on the direction of gravity;

[0069] The accelerometer and gyroscope data of the IMU are integrated to eliminate noise and improve the accuracy of attitude estimation. The formulas for obtaining the pitch angle and roll angle are:

[0070] Pitch angle:

[0071] ;

[0072] Roll angle:

[0073] ;

[0074] in, The angle of rotation of the robot around the X axis; 、 and are the three-axis measurement values ​​of the accelerometer and the acceleration from the I inertial measurement unit; The angle of rotation of the robot around the Y axis, that is, the roll angle;

[0075] Using a 3D pipeline topology model constructed through laser scanning, the central axis of the pipeline is extracted as the ideal trajectory. The actual central axis position of the robot is calculated by combining the IMU attitude angle data and the robot kinematic model. The attitude angles measured by the IMU are converted into the robot's body coordinate system, and the robot's translational displacement is obtained using an odometer or SLAM algorithm. The actual central axis position is compared with the ideal trajectory to calculate the deviation.

[0076] Calculating deviations: Using laser scanning point cloud data, the distance between the robot and the pipe wall is detected in real time to verify the accuracy of the offset calculation. A LiDAR device is used to scan the surrounding environment to obtain a series of point cloud data describing the pipe wall surface features. Each point contains spatial position information relative to the LiDAR coordinate system. For each point or group of point cloud data representing the pipe wall surface, a normal vector is calculated. The robot's precise position, i.e., the coordinates of its center point in space, is determined. A distance calculation formula is used to determine the actual distance from the robot's center point to the pipe wall, selecting one or more pipe wall points from the point cloud data as reference points. The calculated distance is compared with the expected ideal distance.

[0077] The offset is used as input and the drive motor speed is adjusted by the PID controller to make the robot return to the ideal trajectory;

[0078] Finite element analysis of heat conduction:

[0079] Temperature sensors are deployed on the robot's surface to monitor temperature changes within the pipeline in real time. Based on the pipeline's material properties, such as thermal conductivity and specific heat capacity, as well as its geometry, a three-dimensional model of the pipeline is constructed using professional engineering software. The model is meshed, discretizing the continuous pipeline structure into numerous small cells, each representing an independent but interconnected small volume. Appropriate boundary conditions are set based on actual conditions, including but not limited to initial temperature distribution, external heat flux density, and thermal impedance between contact surfaces. The finite element method is used to solve the heat conduction equation and predict the temperature field distribution of various parts of the pipeline over time under given conditions.

[0080] Temperature field distribution:

[0081] Based on the material properties and geometry of the pipeline, a three-dimensional finite element model is constructed and meshed, converting the continuous structure into calculable discrete units. Boundary conditions such as initial temperature, external heat flux, and contact thermal resistance are set according to actual operating conditions to simulate the actual heat transfer environment. The heat conduction equation is solved through numerical calculations to obtain the temperature distribution results of various parts of the pipeline at different times, forming a dynamic temperature field distribution.

[0082] Model the dynamic thermal expansion of the pipe:

[0083] In engineering projects, reservoir water transfer tunnels can experience thermal expansion or contraction due to temperature changes. To ensure the safety and functionality of pipelines during operation, the actual pipe diameter must be dynamically corrected using a formula. The thermal expansion at each location in the pipeline is calculated section by section based on the temperature field distribution.

[0084] For example, the inner wall area expands significantly due to high temperature, while the outer wall area expands less due to low temperature. Materials expand or contract when the temperature changes. When the temperature rises, the pipe will expand radially outward, resulting in an increase in diameter. When the temperature drops, the opposite occurs. The average value of the inner and outer wall temperatures is used as the correction basis to avoid local expansion distortion caused by excessively high or low temperatures on one side.

[0085] Corrected pipe diameter value:

[0086] The original pipe diameter value is obtained through calipers or laser scanning, and the correction value is dynamically adjusted based on real-time feedback, such as pipe deformation and temperature expansion coefficient.

[0087] For example, if the original pipe diameter is 100mm and the diameter expands by 0.5% due to temperature increase, the corrected pipe diameter value is 100.5mm;

[0088] Generate umbrella-like structure telescoping instructions:

[0089] The core of the fuzzy PID controller is to dynamically adjust the three parameters of the traditional PID through fuzzy logic: proportional gain Kp, integral gain Ki, and differential gain Kd. Its input is the error and the error change rate, and the output is the PID parameter correction value. It uses Mamdani fuzzy reasoning to calculate the weight of each rule. The center of gravity method is used to convert the fuzzy output into accurate Kp, Ki, and Kd values. Based on the corrected Kp, Ki, and Kd values, the expansion and contraction amount is adjusted in real time to ensure that the nozzle / sensor always fits the pipe wall. The umbrella-shaped structure is driven by a servo motor or hydraulic cylinder. After receiving the expansion and contraction amount signal, it generates specific instructions and sets the maximum expansion and contraction speed and travel range to avoid mechanical overload.

[0090] Mamdani fuzzy reasoning: The error and error rate of change are converted into linguistic variable values ​​of a fuzzy set, and the degree of membership of each linguistic variable is determined based on a predefined membership function. Based on the fuzzy rule base, the weight of each rule is calculated using the Mamdani reasoning method, and all rule outputs are aggregated to obtain a fuzzy output set. The center of gravity method is used to convert the aggregated fuzzy outputs into precise proportional gain Kp, integral gain Ki, and differential gain Kd values, which are used to adjust the expansion and contraction instructions of the umbrella structure in real time.

[0091] Proportional gain Kp: determines the proportional relationship between the controller output and the current error, that is, the difference between the target value and the actual value;

[0092] Integral gain Ki: Affects the controller output by accumulating all past errors. It is mainly used to eliminate static errors, that is, small errors that still exist when the system tends to be stable.

[0093] Differential gain Kd: Predicts future error trends based on the rate of change of the error, so that measures can be taken in advance to slow down the rate of error change.

[0094] The multi-process collaborative control module dynamically sets the spraying baseline speed and relationship based on the pipe diameter curve and coating thickness requirements. It uses the YOLOv5 model to detect foreign objects on the pipe wall in real time. When the obstacle height is greater than the height threshold c, it pauses the process, triggers removal, and continues the process. If removal is still not possible, the location is marked and the control console is notified for manual intervention.

[0095] When the obstacle height is identified to be less than or equal to the height threshold c, the vehicle is stopped and local grinding is triggered;

[0096] Dynamically set the spraying base speed:

[0097] By determining the baseline spraying speed and based on the principle of paint flow conservation, precise control is achieved for the dynamic balance between pipe diameter, coating thickness, and paint flow. Laser scanning is used to obtain the real-time pipe diameter, which, combined with temperature sensor data, calculates the thermal expansion of the pipeline to obtain the corrected actual pipe diameter. The effective flow rate is adjusted according to the ambient temperature to offset changes in paint fluidity caused by high temperatures.

[0098] Based on the characteristics of the selected coating, such as solid content percentage and viscosity, calculate the volume of coating required to achieve a specified thickness per unit area. Based on the output efficiency of the spraying equipment under different conditions and the length of the pipeline, set a reasonable spraying speed so that the coating can evenly cover the pipeline surface.

[0099] Real-time detection of foreign matter on pipe walls:

[0100] In YOLOv5, the C2f module replaces the original C3 module to enhance the ability to extract gradient flow information. The C2f module uses a residual structure and 1×1 convolution to reduce dimensionality, reducing computational complexity while retaining more detailed features. This replaces the traditional CIoU loss function, solves the problem of inaccurate positioning of small objects with adhesion, and improves the bounding box regression accuracy. Channel attention or effective channel attention is introduced in Backbone or Neck to enhance the focus on key features of foreign objects and suppress background interference.

[0101] To address issues such as uneven lighting and noise interference that may exist within pipelines, the following optimization techniques are required: Contrast-limited adaptive histogram equalization is used to enhance image contrast and reduce the impact of ambient noise on detection. BiFPN or FPN is introduced in the neck region to achieve cross-scale feature fusion and adapt to foreign object detection of different sizes. YOLOv5 is used to reduce the number of parameters and computational complexity, adapting to real-time operation on edge devices.

[0102] For example, the robot's visual sensor inside the pipeline collects images of the pipe wall in real time. CLAHE is performed on the image to improve contrast in low-light areas and remove noise and background interference. Mosaic data enhancement and adaptive anchor box calculation are used to improve the model's generalization ability for foreign objects, adapt the input image size, and ensure that small targets can be effectively detected. The detected foreign object category and bounding box coordinates are output, along with confidence and category probability.

[0103] If the pipe inner wall diameter and camera parameters are known, the obstacle height can be calculated through trigonometric relationships. Based on past experiments showing the impact of obstacles on spraying, a pre-set obstacle height threshold c is set to trigger the early warning system. The location and height of foreign objects can also be marked on the monitoring screen to assist operations and maintenance personnel in making quick decisions.

[0104] Long-term water erosion in a reservoir's water supply tunnel has caused scaling on the pipe walls, necessitating real-time monitoring of scale thickness to prevent blockage. A camera mounted on a robot collects images every 5 meters. The YOLOv5 model is trained on a dataset of scaling samples, optimizing the NWD loss function to improve small object detection accuracy. The CBAM attention mechanism is introduced to suppress interference from water reflections, and the scale height is calculated using a geometric formula. When the scale height exceeds 5mm, the system automatically triggers a manual intervention prompt and records the location coordinates for subsequent processing.

[0105] Trigger clear:

[0106] Use a grinding and rust removal vehicle to grind away foreign matter, and use a vacuum cleaning vehicle to clean up the residue after grinding.

[0107] The coating thickness closed-loop control module obtains thickness feedback in real time, compares it with the target value, adjusts the spray motor speed, and performs floating compensation on the nozzle;

[0108] During the corrosion protection process of industrial pipelines, the uniformity and accuracy of coating thickness directly impact the subsequent quality of the pipeline. Traditional open spray systems rely on preset parameters and are difficult to cope with environmental fluctuations, resulting in coating thickness deviations. High-precision and highly consistent coating control is achieved through real-time feedback, dynamic adjustment, and floating compensation. Electromagnetic eddy currents are used to automatically detect coating thickness, and the operator sets the target coating thickness through the HMI. If the measured value is lower than the target, the spray volume is increased; if it is higher, the spray volume is reduced. The motor speed is dynamically adjusted based on the deviation to eliminate long-term deviations, prevent the coating thickness from continuously deviating from the target, predict deviation trends, and suppress oscillations caused by fluctuations in workpiece movement speed.

[0109] Floating compensation for the print head:

[0110] By adjusting the nozzle posture, the paint is ensured to be evenly covered. The integrated force sensor and displacement sensor detect the contact force and relative position between the nozzle and the workpiece in real time. For example, when the inner surface of the pipe is raised, the sensor detects the increase in contact force and triggers the nozzle to deflect upward. The floating resistance of the nozzle is automatically adjusted according to the contact force to avoid excessive oscillation or lag. YOLOv5 is used to detect uneven spraying on the pipe wall. The closed-loop system automatically increases the local coating thickness and adjusts the nozzle angle through floating compensation to cover irregular areas.

[0111] Example 2

[0112] See also Figure 2 Based on Example 1, Example 2 of the present application further provides a method for controlling a pipe polishing robot, comprising the following specific steps:

[0113] Step 1: Obtain the pipe diameter curve through continuous laser scanning, and use the SLAM algorithm to stitch the continuously scanned point cloud data in real time to construct a dynamic three-dimensional topological model of the pipeline; obtain the robot's posture data and calculate the offset of the robot's central axis in the pipeline; obtain the monitored ambient temperature in the pipeline, and use heat conduction finite element analysis to construct a dynamic thermal expansion model of the pipeline, and correct the pipe diameter value according to the thermal expansion model; based on the offset and the corrected pipe diameter value, calculate the expansion and contraction of the umbrella structure through a fuzzy PID controller, and generate the expansion and contraction instructions of the umbrella structure;

[0114] Step 2: Dynamically set the spraying baseline speed and relationship based on the pipe diameter curve and coating thickness requirements; use the YOLOv5 model to detect foreign objects on the pipe wall in real time. When the obstacle height is greater than the height threshold c, the process is paused, the removal is triggered, and the process continues; if the obstacle still cannot be removed, the location is marked and the control console is notified for manual intervention;

[0115] When the obstacle height is identified to be less than or equal to the height threshold c, the vehicle is stopped and local grinding is triggered;

[0116] Step 3: Get thickness feedback in real time, compare it with the target value, adjust the spray motor speed, and perform floating compensation on the nozzle.

[0117] The above embodiments can be implemented in whole or in part by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented in whole or in part in the form of a computer program product. Those skilled in the art will appreciate that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution.

[0118] The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, and may be located in one place or distributed across multiple network units. Some or all of these units may be selected to achieve the purpose of this embodiment as needed.

[0119] The above is only a specific implementation method of the present application, but the scope of protection of the present application is not limited thereto. Any technician familiar with this technical field can easily think of changes or replacements within the technical scope disclosed in this application, which should be covered by the scope of protection of the present application.

Claims

1. A system for controlling a pipe polishing robot, characterized in that: The system includes: The variable diameter adaptive preprocessing module acquires the pipe diameter curve through continuous laser scanning, and uses the SLAM algorithm to stitch the continuously scanned point cloud data in real time to construct a dynamic three-dimensional topological model of the pipeline. It also acquires the robot's posture data and calculates the offset of the robot's central axis within the pipeline. It obtains the monitored ambient temperature within the pipeline and constructs a dynamic thermal expansion model of the pipeline through heat conduction finite element analysis. The pipe diameter value is corrected according to the thermal expansion model. Based on the offset and the corrected pipe diameter value, the fuzzy PID controller calculates the expansion and contraction of the umbrella structure and generates the expansion and contraction instructions for the umbrella structure. The process of real-time stitching by SLAM algorithm is as follows: Using the LIO-SLAM algorithm, dynamic point cloud data stitching is achieved through the following steps: S101: Use IMU data, accelerometer and gyroscope to calculate the robot's posture changes in real time; S102: Aligning point cloud data of adjacent frames using the ICP algorithm; S103: Based on the SLAM framework, the spliced ​​point cloud data is constructed into a dynamic 3D topological model of the pipeline; Voxel filtering is used to reduce the point cloud density while preserving geometric features; statistical outlier detection is used to remove noise points; The process of constructing the dynamic three-dimensional topological model of the pipeline is as follows: Based on dense point cloud data, the entire pipe cross section is covered to generate the inner wall of the complex shape of the pipe. Each frame of point cloud data contains spatial point coordinate information, and the point cloud is filtered to remove isolated points. Using the ICP algorithm, the newly acquired point cloud frame is matched with the previously constructed map to find the optimal rigid body transformation between the two frames of point cloud. By integrating the results of multiple local registrations, a consistent global map is formed. Based on the registered point cloud data, computer vision algorithms are used to extract key geometric features inside the pipeline and build a detailed topological structure. As the robot continues to advance, new point cloud data is continuously added to the existing map, enabling dynamic updates to the pipeline's 3D model. The multi-process collaborative control module dynamically sets the spraying baseline speed and relationship based on the pipe diameter curve and coating thickness requirements. It uses the YOLOv5 model to detect foreign objects on the pipe wall in real time. When the obstacle height is greater than the height threshold c, it pauses the process, triggers removal, and continues the process. If removal is still not possible, the location is marked and the control console is notified for manual intervention. When the obstacle height is identified to be less than or equal to the height threshold c, the vehicle is stopped and local grinding is triggered; The coating thickness closed-loop control module obtains thickness feedback in real time, compares it with the target value, adjusts the spray motor speed, and performs floating compensation on the nozzle; The process of correcting the pipe diameter value is as follows: Based on the material properties of the pipeline, the model is meshed, discretizing the continuous pipeline structure into many small cells. Appropriate boundary conditions are set according to the actual situation, and the finite element method is used to solve the heat conduction equation to predict the temperature field distribution of each part of the pipeline over time under given conditions. According to the temperature field distribution, the thermal expansion of each position of the pipeline is calculated section by section; the average value of the inner and outer wall temperatures is used as the correction basis; The specific process of generating the umbrella structure telescoping instruction is as follows: The three parameters of traditional PID, namely proportional gain Kp, integral gain Ki, and differential gain Kd, are dynamically adjusted through fuzzy logic. The input is the error and the error change rate, and the output is the PID parameter correction value. Mamdani fuzzy reasoning is used to calculate the weight of each rule, and the center of gravity method is used to convert the fuzzy output into accurate Kp, Ki, and Kd values. The expansion and contraction amount is adjusted in real time based on the corrected Kp, Ki, and Kd values.

2. A system for controlling a pipe polishing robot according to claim 1, characterized in that: The process of obtaining the pipe diameter curve is as follows: Perform arc fitting on the spliced ​​point cloud data to calculate the inner diameter value of each cross section; identify the diameter-changing area through curvature calculation and generate a segmented continuous pipe diameter curve.

3. The system for controlling a pipeline polishing robot according to claim 1, characterized in that: The process of constructing the dynamic three-dimensional topological model of the pipeline is as follows: Generate the inner wall of the pipeline based on the point cloud data; each frame of point cloud data contains spatial point coordinate information, and the point cloud is filtered to remove isolated points; Using the ICP algorithm, a newly acquired point cloud frame is matched with the previously constructed map to find the rigid body transformation between the two frames of point clouds. By integrating the results of multiple local registrations, a global map is formed. Based on the registered point cloud data, computer vision algorithms are used to extract key geometric features inside the pipeline and dynamically update them.

4. The system for controlling a pipeline polishing robot according to claim 1, characterized in that: The process of calculating the offset of the robot's central axis in the pipeline is as follows: The robot's angular velocity and acceleration are acquired through sensors, and the central axis of the pipeline is extracted as the ideal trajectory based on the three-dimensional topological model of the pipeline constructed by laser scanning. The actual central axis position of the robot is calculated by combining the attitude angle data of the IMU and the robot's kinematic model. The attitude angle measured by the IMU is converted into the position in the robot's coordinate system, and the robot's translational displacement is acquired through the odometer. The actual central axis position is compared with the ideal trajectory, and the deviation is calculated.

5. The system for controlling a pipeline polishing robot according to claim 1, characterized in that: The process of dynamically setting the spraying reference speed and relationship is as follows: By determining the spraying baseline speed and based on the principle of paint flow conservation, precise control is achieved for the dynamic balance between pipe diameter, coating thickness, and paint flow. Laser scanning is used to obtain the real-time pipe diameter, which, combined with temperature sensor data, calculates the thermal expansion of the pipeline to obtain the corrected actual pipe diameter. The effective flow rate is adjusted according to the ambient temperature to offset changes in paint fluidity caused by high temperatures. Set the spraying speed according to the output efficiency of the spraying equipment under different conditions.

6. The system for controlling a pipeline polishing robot according to claim 1, characterized in that: The process of floating compensation of the nozzle is as follows: Electromagnetic eddy currents are used to automatically detect coating thickness, and the operator sets the target coating thickness through the HMI. The motor speed is dynamically adjusted based on the deviation to eliminate long-term deviations. Integrated force sensors and displacement sensors detect the contact force and relative position between the printhead and the workpiece in real time. The floating resistance of the printhead is automatically adjusted based on the contact force, and the printhead angle is adjusted through floating compensation.

7. A method for controlling a pipeline polishing robot, based on the system for controlling a pipeline polishing robot according to any one of claims 1 to 6, characterized in that: The steps include: Step 1: Obtain the pipe diameter curve through continuous laser scanning, and use the SLAM algorithm to stitch the continuously scanned point cloud data in real time to construct a dynamic three-dimensional topological model of the pipeline; obtain the robot's posture data and calculate the offset of the robot's central axis in the pipeline; obtain the monitored ambient temperature in the pipeline, and use heat conduction finite element analysis to construct a dynamic thermal expansion model of the pipeline, and correct the pipe diameter value according to the thermal expansion model; based on the offset and the corrected pipe diameter value, calculate the expansion and contraction of the umbrella structure through a fuzzy PID controller, and generate an expansion and contraction instruction for the umbrella structure; Step 2: Dynamically set the spraying baseline speed and relationship based on the pipe diameter curve and coating thickness requirements; use the YOLOv5 model to detect foreign objects on the pipe wall in real time. When the obstacle height is greater than the height threshold c, the process is paused, the removal is triggered, and the process continues; if the obstacle still cannot be removed, the location is marked and the control console is notified for manual intervention; When the obstacle height is identified to be less than or equal to the height threshold c, the vehicle is stopped and local grinding is triggered; Step 3: Get thickness feedback in real time, compare it with the target value, adjust the spray motor speed, and perform floating compensation on the nozzle.

Images